Fig. 3: Comparative segmentation analysis.
From: Achieving flexible fairness metrics in federated medical imaging
We evaluate FlexFair against baseline methods on segmentation tasks across three datasets: cervical cancer, polyp, and fundus vascular. Violin-box plots depict the distribution of the top 20 test results for each method across different weight configurations and random seeds. The boxes represent the interquartile range (IQR), with the median marked by the white line and the mean indicated by the red 'x'. Whiskers extend to data points within 1.5 times the IQR, with black diamonds showing outliers. Dice scores, serving as a metric for segmentation accuracy, highlight FlexFair’s consistently superior performance across all tasks, characterized by a tighter distribution around higher median and mean values compared to the baseline methods. Source data are provided as a Source Data file.
